DeepMark embeds inaudible identity into AI voice agents

What's the deal? DeepMarkDealroom has a profile for this one. Try Dealroom → (YC F26)has launched, embedding an inaudible, machine-readable identity directly into AI-generated media, so any receiver can confirm whether a call came from an authorized AI agent. The company was announced on X on 22 September 2026.

How it works: A four-step flow powers both pillars. Register: enroll an identity, an authorized voice agent, a model, a provider or a content provenance record,and get a credential to embed. Embed at generation: DeepMark weaves an imperceptible signal into audio, image or video as it is produced. Detect or verify later:in a call, an API, enterprise infrastructure or the network,the signal is detected and its payload resolved,with clear pass-fail results, from as little as a few seconds of audio. Resolve and prove:the payload resolves to "this is AI-generated and marked" or "this voice is an authorized agent," with signed evidence.

Why it matters: The product sits at the intersection of two regulatory and fraud pressures. EU AI Act Article 50 transparency obligations and the Code of Practice's multi-layer approach push for machine-readable marking of AI-generated content. Meanwhile voice-agent fraud keeps rising, particularly in banking, contact centers, healthcare and insurance, where scammers impersonate institutions. DeepMark's payload carries up to 128 bits, enough for an agent or session identity, model or provider ID, or a durable pointer to a signed certificate,and it says it is built to survive VoIP, transcoding, codecs and compression.

Signal: DeepMark is YC-backed (F26)and cites academic work on watermarking and audio forensics. Its positioning is compliance-first, aimed at enterprises that must prove provenance or authorize live AI voice agents. If adoption spreads at platform level, verification could become a default expectation for AI voice interactions, rather than a niche audit feature.

Read more: Y Combinator on X · DeepMark